LLMs default to Markdown due to training data prevalence

Large language models like ChatGPT, Claude, and Gemini often default to using Markdown for formatting their responses because this markup language was prevalent in their training data. Markdown, originally designed for bloggers to easily convert plain text to HTML, has become a standard in developer communities through platforms like GitHub and tools like Obsidian. Its efficiency in structuring text with minimal markup makes it ideal for both human readability and machine parsing, leading models to generate it naturally without explicit instructions. Attempts to instruct models to avoid Markdown can result in the disappearance of headings and lists, highlighting the format's deep integration into their output. AI

IMPACT LLMs' natural use of Markdown streamlines output for developers and users, reducing token usage and improving readability in chat interfaces.

RANK_REASON The item discusses the technical reasons behind LLM output formatting, drawing on historical context and training data analysis, rather than announcing a new release or event.

Read on dev.to — LLM tag →

• Aaron Swartz
• arXiv
• ChatGPT
• Claude
• Cloudflare
• CommonMark
• EleutherAI
• Gemini
• GitHub
• GitHub Flavored Markdown
• John Gruber
• Markdown
• Obsidian
• Jupyter
• Stack Exchange
• The Pile

• product
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →